EP0314370B1 - Bounded lag distributed discrete event simulation method and apparatus - Google Patents
Bounded lag distributed discrete event simulation method and apparatus Download PDFInfo
- Publication number
- EP0314370B1 EP0314370B1 EP88309768A EP88309768A EP0314370B1 EP 0314370 B1 EP0314370 B1 EP 0314370B1 EP 88309768 A EP88309768 A EP 88309768A EP 88309768 A EP88309768 A EP 88309768A EP 0314370 B1 EP0314370 B1 EP 0314370B1
- Authority
- EP
- European Patent Office
- Prior art keywords
- event
- subsystem
- simulation
- time
- events
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Lifetime
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Definitions
- This invention relates to methods for simulating behaviour of a system and to systems for performing discrete event simulation.
- Computer simulation has become very important in recent years because of the many applications where simulation of systems is highly beneficial.
- One such application is the use of simulation in the design of complex systems. These may be electronic systems such as a telecommunications switching network, robot based flexible manufacturing systems, process control systems, health care delivery systems, transportation systems, and the like. Design verification through simulation plays an important role in speeding up the design and insuring that it conforms to the specification.
- Another application is the use of simulation in analyzing, and tracking down, faults appearing in operating systems.
- Still another application is optimizing the operation of existing systems through repeated simulations, e.g., the operation of a manufacturing facility, the operation of the telecommunications network, scheduling and dispatching, etc.
- Yet another application is the use of simulation to predict the operation of systems which for various reasons can not be tested (e.g., response to catastrophe).
- Simulations can be classified into three types: continuous time, discrete time, and discrete event.
- Discrete event simulation means simulation of a system in which phenomena of interest change value or state at discrete moments of time, and no changes occur except in response to an applied stimulus. For example, a bus traveling a prescribed route defines a discrete event system in which the number of passengers can change only when the bus arrives at a bus stop along the route.
- a number of parallel processors form a simulation multicomputer network, and it is the entire network that is devoted to a simulation task. More specifically, each processor within the network is devoted to a specific portion of the system that is simulated; it maintains its own event list and communicates event occurrences to appropriate neighbor processors. Stated conversely, if one views a simulated system as a network of interacting subsystems, distributed simulation maps each subsystem onto a processor of the multi computer network.
- the time stamping is required, of course, to maintain causality so that in a message-receiving node an event that is scheduled for time T is not simulated when other incoming messages can still arrive with a time-stamp of less than T. Because of this, when a particular node is able to receive input from two sender nodes, it cannot simulate an event with any assurance that it would not be called upon to refrain from simulating the event, until it receives a message from both sender nodes. Waiting to receive a message from all inputs slows the simulation process down substantially and can easily result in a deadlock cycle where each node waits for a previous node, which amounts to the situation of a node waiting for itself.
- the null message approach expends computing resources in generating, sending, and reading the null messages; the recovery approach expends computing resources to detect and recover from a deadlock, and roll-back approach expends computing resources in simulating events and then undoing the work that was previously done.
- a simulation system is rcalized that avoids all blocking and advances the simulation time in an efficient manner.
- the efficiency is achieved by each node independently evaluating for itself a time interval that is not "at risk” and simulating all events that are within that evaluated time.
- a point in time is not "at risk” for a considered node if no other node in the system has a scheduled event that can affect the simulation at the considered node.
- FIG. 1 presents a pictorial explanation that may aid in understanding the invention.
- vertical lines 21, 22, 23, 24, 25, 26 and 27 represent seven simulation nodes and their simulation time lines (where time advances upward).
- the circles along these lines (30-43) represent events that have been, or are scheduled to be processed (i.e., simulated). These events may cause change of value or state, i.e., other events, at the originating node or at some other nodes.
- node 24 is the node of concern, but it is understood that the consideration undertaken by node 24 are concurrently taken by all other nodes.
- the horizontal distances between line 24 and the other lines represent the time delays for events in other nodes to affect the value or state at node 24. Accordingly, event 30 processed at node 21 for time T1 may cause an event 40 at node 24 to be scheduled for some time not earlier than T2, as shown in FIG. 1.
- the interval between T1 and T2 equals the delay between line 24 and line 21 (i.e., the horizontal distance between the lines).
- the events depicted in FIG. 1 can be divided into two groups: those that have been simulated (and marked with a cross), and those that are yet to be simulated (un-crossed).
- the simulated events need not be considered for simulation because their effects are already represented by those events which have not been simulated (for example, event 30 has caused event 40; the first has been simulated, while the second is yet to be considered for processing).
- event 33 in FIG. 1 is earliest in time from among all of the nodes.
- the time of event 33 forms the current floor simulation time of the system.
- the floor is depicted in FIG. 1 by dashed line 45.
- a time interval beginning with the floor simulation time is selected for consideration.
- This time interval which I call the bounded lag interval, can be a convenient time interval that takes into account the number of nodes in the system, the number of events to be simulated, and the computing power of the processors employed. All events scheduled within the bounded lag interval can be affected by events scheduled at other nodes within the bounded lag time interval, but only if those nodes are at a time distance from the affected node 24 that is less than the selected bounded lag interval. That reduces the number of nodes that need to be considered. In the FIG.
- the bounded lag interval ends at dashed line 46; and as drawn, the nodes that need to be considered for their potential effect on node 24 are nodes 22, 23, 25 and 26. Nodes 21 and 27 are outside the bounded lag delay and their scheduled events within (or outside) the bounded lag need not be considered.
- next scheduled event is event 35, and as drawn, only nodes 23 and 25 need to be considered.
- event 35 In determining whether event 35 is to be simulated, one can observe that only event 34 is scheduled early enough to have a possible impact on event 35. That event can be analyzed and if it is determined that it does not affect event 35, then event 35 can be simulated. Alternatively, it may prove even more efficient to refrain from processing event 35 simply because of the potential effect by event 34. In the following description, this alternative approach is taken because it saves the process of evaluating what event 34 may do.
- FIG. 2 presents a block diagram of a concurrent event simulator. It comprises a plurality of node controllers 51 that are connected to a communications and common processing network 52. Network 52 performs the communication and synchronization functions, and in some applications it also performs some of the computations that are needed by the entire system.
- the node controllers can be realized with conventional stored program computers that may or may not be identical to each other.
- Each of the node controllers which corresponds to a node in the FIG. 1 depiction, is charged with the responsibility to simulate a preassigned subsystem of the simulated system.
- Each controller C i maintains an event list ⁇ i that is executed by simulating each event in the list in strict adherence to the scheduled event simulation times.
- the bounded lag interval is selectively fixed, and that in conformance with the selected bounded lag interval each controller 51 is cognizant of the processors with which it must interact to determine whether events are scheduled.
- the process by which the event simulations are enabled is carried out by the controllers in accordance with the above-described principles, as shown, for example, by the flowchart of FIG. 3.
- Block 100 is the flow control block. It tests whether the floor simulation time, T floor , is less than the end of the simulation time, T end . As long as T floor ⁇ T end , the simulation continues by passing control to block 110. When T floor reaches or exceeds T end , the simulation ends. Block 110 determines the floor simulation time of the simulated system at each iteration. That is, block 110 determines the lowest event time of the scheduled events dispersed among the event lists ( ⁇ i ) of controllers 51 (C i ) that are are yet to be simulated.
- block 110 evaluates the floor simulation time in accordance with the equation where N is the total number of node controllers 51 and T i is the time of the event, e i , which has the earliest scheduled time among the events e' in the event list ⁇ i ; i.e., Block 110 can be implemented in each of the controllers by having each controller broadcast to all other controllers (via network 52) its T i and evaluate for itself the minimum T i which defines T floor . Alternatively, block 110 can be implemented within communications and common processing network 52 by having controllers 51 send their T i values to network 52, and having network 52 select the minimum T i and broadcast it as T floor back to the controllers.
- each of the controllers evaluates its earliest “at risk” time. This is accomplished by network 52 distributing the T i information to neighboring controllers, as required, and each controller C i evaluates the "at risk” demarcation point, ⁇ i , from the T i information.
- This "at risk” point is defined as the earliest time at which changes at the neighboring controllers can affect the history simulated by the controller, based on the neighboring controllers' own scheduled events or based on a response to an event from the controller itself (reflection). This is expressed by the following equation:
- processor C i simulates all or some of the scheduled events whose times are earlier that ⁇ i .
- the time T i is advanced with each simulation of an event, and the simulated event is deleted from ⁇ i .
- new events are called to be scheduled, then those events are sent to network 52 for transmission to the appropriate node controllers.
- the execution of events is called to be blocked, that information is also sent to network 52, and thereafter to the appropriate node controllers for modification of the event lists.
- node controller 22 Since there are no events scheduled for node controller 24 between the time of T floor , and point 50, no progress in simulations is made by this controller. Concurrently, node controller 22 simulates event 33 (since it is positioned at T floor , and no other event can affect it), and node 25 simulates event 34 since neither node 26 nor node 24 (the closest nodes) have any events scheduled prior to the time of event 34. Node 27 probably also simulates event 36, but this is not certain, since FIG. 1 does not show all of the neighbors of node controller 27.
- event 39 is simulated, event 37 at node 27 is simulated, but at node 26 event 38 is not simulated because it is beyond the "at risk" point of node 26 caused by the position of event 37 at node 27.
- the next T floor moves to the time of event 38, and the process continues to simulate additional events.
- neighbors(i) refers to nodes that communicate directly with node i.
- the auxiliary variable ⁇ k i represents an estimate of the earliest time when events can affect node i after traversing exactly k links. It can be shown that the iteration test is always met within a relatively low number of steps, depending on the value of the bounded lag interval, B. To account for opaque periods, the evaluation of ⁇ is augmented to be where op ji is the end of opaque period (when communication unblocks) for node j in the direction of node i.
- each ⁇ i reduces to computing the minimum of the opaque periods relative to block i; to wit:
- T floor increases because the event determining that value can always be simulated.
- the movement of T floor is affected, however, by how closely the nodes are separated and the scheduled events.
- One other observation that can be made is that the above-described procedure is very conservative, in the sense that an assumption is made that whenever an event is scheduled to be simulated at one node controller, it will always have an effect on the neighboring controller (after the appropriate delay).
- Knowledge that an event scheduled for simulation will not affect a neighboring node can be put to use to simulate more events concurrently (have fewer node controllers that are idle). This can be accomplished by communicating not only the T i of the earliest scheduled event in each list, but also the effect that it may have on neighboring node controllers. If fact, each list can broadcast more than just the earliest scheduled event.
- the design decision that a practitioner must make, of course, is whether fewer iterations but more complex evaluations are economically justifiable.
- FIG. 4 presents a block diagram of one realization for node controller 51.
- this embodiment relates to use of the iterative method for evaluating ⁇ (with the use of the auxiliary variable ⁇ ) when the delays between changes in one subsystem and the effect of those changes on other subsystems are not insignificant, it will be appreciated by the skilled artisan that the other realizations for computing ⁇ i are substantially similar to this realization. It will also be appreciated that although node controller 51 is shown in FIG. 2 as one of a plurality of controllers, such plurality can be realized within a single processor of sufficient computing power.
- state register 56 defines the current state of the simulated subsystem and event list 53 specifies the events that are scheduled to be simulated.
- Processor 54 is the event simulation processor, and it is responsive to state register 56, to event list 53 and to register 58.
- Register 58 is the ⁇ register and, as the name implies, it stores the value of ⁇ for the controller. Based on the value of ⁇ and the scheduled time of the event at the top of the event list, processor 54 simulates the event in conformance with the state of the subsystem and develops a new state of the subsystem as well as, perhaps, new events.
- the new state is stored in register 56, new events scheduled for the controller are stored in event list 53 via line 61, and events affecting other controllers are communicated out via line 62. Events produced at other controllers that may affect this controller are accepted via line 63 and stored in event list 53 through processor 54.
- processor 54 is the event simulation processor
- processor 55 is the synchronization processor.
- Processor 54 is shown in FIG. 4 as a separate processor but in practice a single processor may serve the function of both processor 54 and processor 55.
- Processor 55 receives information from event list 53 concerning the time of the event in list 53 that possesses the earliest simulation time. It transmits that information to other controllers via line 64 and receives like information from all other relevant nodes via line 65. From that information processor 55 develops the value of T floor and stores that value in register 57.
- Processor 55 also receives ⁇ j information from its neighbor controllers (controllers where changes can affect the controller directly) via line 66, and transmits its own ⁇ i values via line 67. With the aid of T floor and the other incoming information, processor 55 performs the iterative computations to develop the values of ⁇ k i and ⁇ k i . Those values are stored by processor 55 in registers 57 and 58, respectively.
- the above assigns the task of computing T floor to processor 55.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Multi Processors (AREA)
Description
- This invention relates to methods for simulating behaviour of a system and to systems for performing discrete event simulation.
- Computer simulation has become very important in recent years because of the many applications where simulation of systems is highly beneficial. One such application is the use of simulation in the design of complex systems. These may be electronic systems such as a telecommunications switching network, robot based flexible manufacturing systems, process control systems, health care delivery systems, transportation systems, and the like. Design verification through simulation plays an important role in speeding up the design and insuring that it conforms to the specification. Another application is the use of simulation in analyzing, and tracking down, faults appearing in operating systems. Still another application is optimizing the operation of existing systems through repeated simulations, e.g., the operation of a manufacturing facility, the operation of the telecommunications network, scheduling and dispatching, etc. Yet another application is the use of simulation to predict the operation of systems which for various reasons can not be tested (e.g., response to catastrophe).
- Simulations can be classified into three types: continuous time, discrete time, and discrete event. Discrete event simulation means simulation of a system in which phenomena of interest change value or state at discrete moments of time, and no changes occur except in response to an applied stimulus. For example, a bus traveling a prescribed route defines a discrete event system in which the number of passengers can change only when the bus arrives at a bus stop along the route.
- Of the three simulation classes, from computation standpoint it appears that discrete event simulation is potentially the least burdensome approach because simulation of time when nothing happens is dispensed with. Of course, synchronization of the event simulations must be considered when parallelism is employed. Most often, a discrete event simulator progresses by operating on an event list. An event at the top of the list is processed, possibly adding events to the list in the course of processing, and the simulation time is advanced. Thereafter, the processed event at the top of the list is removed. This technique limits the speed of simulation to the rate at which a single processor is able to consider the events one at a time. In a parallel scheme many processors simultaneously are engaged in the task creating a potential for speeding up the simulation. Although techniques for performing event list manipulation and event simulation in parallel have been suggested, large scale performance improvements are achieved only by eliminating the event list in its traditional form. This is accomplished by distributed simulations.
- In a distributed simulation, a number of parallel processors form a simulation multicomputer network, and it is the entire network that is devoted to a simulation task. More specifically, each processor within the network is devoted to a specific portion of the system that is simulated; it maintains its own event list and communicates event occurrences to appropriate neighbor processors. Stated conversely, if one views a simulated system as a network of interacting subsystems, distributed simulation maps each subsystem onto a processor of the multi computer network.
- Although distributed simulation provides parallelism which has the potential for improving the simulation speed, allocation and synchronization of work among the processors is a major concern which may impede the realization of the improvement goals. One well known approach for distributed simulation has been proposed by Chandy and Misra in "Distributed Simulation: A Case Study in Design and Verification of Distributed Programs,"IEEE Transactions on Software Engineering, Vol. SE-5, No. 5, September 1979, pp. 440-452, and by Chandy, Holmes and Misra in "Distributed Simulation of Networks,"Computer Networks, Vol. 3, No. 1, February 1979, pp. 105-113. In this approach, they recognize that physical systems to be simulated are composed of independent but interacting entities, and that those entities should be mapped onto a topologically equivalent system of logical nodes. Interaction between nodes is accomplished by the exchange of time-stamped messages which include the desired message information and identify the simulation time of the sending node. In accordance with the Chandy-Holmes-Misra approach, the nodes interact only via messages. There are no global shared variables, each node is activated only in response to a message, each node maintains its own clock, and finally, the time-stamps of the messages generated by each node are non-decreasing (in time). In this arrangement, each of the nodes works independently to process the events assigned to it in the correct simulated order. Thus, independent events can be simulated in parallel, within different nodes, even if they occur at different simulated times.
- The time stamping is required, of course, to maintain causality so that in a message-receiving node an event that is scheduled for time T is not simulated when other incoming messages can still arrive with a time-stamp of less than T. Because of this, when a particular node is able to receive input from two sender nodes, it cannot simulate an event with any assurance that it would not be called upon to refrain from simulating the event, until it receives a message from both sender nodes. Waiting to receive a message from all inputs slows the simulation process down substantially and can easily result in a deadlock cycle where each node waits for a previous node, which amounts to the situation of a node waiting for itself.
- To remedy the wait problem, artisans have been employing recovery and avoidance techniques. In the recovery technique, proposed by Chandy and Misra, upon detecting a deadlock, the processors in the network exchange messages in order to determine which of the waiting nodes can process their events in spite of the apparent deadlock. This is described in K. M. Chandy and J. Misra, "Asynchronous Distributed Simulation via a Sequence of Parallel Computations,"Communications of the ACM, Vol. 24, No. 4, April 1981, pp. 198-206. In the avoidance technique, on the other hand, certain types of nodes send null messages under specific conditions even when no instructions for other nodes are called for. By this technique, nodes can be advanced more quickly in their simulation time. Jefferson and Sowizral, in "Fast Concurrent Simulation Using the Time Warp Mechanism,"Distributed Simulation, 1985, The Society for Computer Simulation Multiconference, San Diego, California, suggest a different technique where each node is allowed to advance in its simulation time "at its own risk," but when a message arrives that would have caused some events to not have been simulated, then a "roll-back" is executed to undo the simulation that was done. Roll-back of a node may not be difficult, perhaps, but the fact that the simulated event(s) that need to be rolled back may have caused messages to be sent to other nodes does complicate the task substantially. To achieve the roll-back, Jefferson et al. suggest the use of "anti-messages," which are messages that parallel the original messages, except that they cause the performance of some action that "undoes" the original action.
- Neither of these techniques is very good because each potentially expands an inordinate amount of computation time in making sure that the overall simulation advances properly. The null message approach expends computing resources in generating, sending, and reading the null messages; the recovery approach expends computing resources to detect and recover from a deadlock, and roll-back approach expends computing resources in simulating events and then undoing the work that was previously done.
- 1986 WINTER SIMULATION CONF. PROC., Washington, DC, 8th - 10th December 1986, pages 417-423, IEEE, New York, US; D.W. JONES: "Concurrent simulation: An alternative to distributed simulation" discloses a new approach to parallel simulation where previous work in this area has concentrated on distributed simulation. The new approach uses spatial decomposition to allow simulations to be run on networks of machines, where the message flow between processors in the network is related closely to the topology of the system being simulated. The disclosed approach, concurrent simulation, is based on temporal decomposition. This allows natural use to be made of the shared memory facilities and load-balancing capabilities of multiprocessors.
- According to one aspect of this invention there is provided a method as claimed in
claim 1. - According to another aspect of this invention there is provided a system as claimed in claim 14.
- Recognizing that in physical systems there is always some delay between the time when one part of the system does something, and the time when another part of the system realizes that something was done, a simulation system is rcalized that avoids all blocking and advances the simulation time in an efficient manner. The efficiency is achieved by each node independently evaluating for itself a time interval that is not "at risk" and simulating all events that are within that evaluated time. A point in time is not "at risk" for a considered node if no other node in the system has a scheduled event that can affect the simulation at the considered node. By also restricting the simulation of scheduled events at any one time to a chosen simulated time segment (bounded lag) beginning with the lowest simulation time found among the nodes, allows the evaluation of the "at risk" time interval to be based on only a subset of the nodes that can potentially affect the simulation at the considered node. This simplification results from the the fact that there are delays between nodes, and that the lower bounds for those delays are fixed and known apriorily.
- In simulating systems where some nodes affect other nodes only through intermediate nodes, opaque periods can be experienced when, because of the specific process that is being simulated, such an intermediate node "promises" that a particular route emanating from this node would be busy for a set period of time, and thereby also "promises" that no other node can use this route as a conduit to affect other nodes. That, in effect, increases the propagation delay from the nodes that use the busy intermediate route, which, in turn, increases the allowance for the simulation time within the bounded lag that is not "at risk".
-
- FIG. 1 illustrates the timing inter-relationships of events processed in a multi-processor environment with recognized delay between events and their effect on other events;
- FIG. 2 depicts a block diagram of a multi-processor arrangement for simulating events ;
- FIG. 3 presents a flow chart describing the steps carried out in each of the processors of FIG. 2 in the course of event simulation; and
- FIG. 4 describes one realization for the node controllers of FIG. 2.
- As indicated above, one of the major problems with the prior art distributed simulation systems is their failure to realize and take advantage of the fact that a delay is always present between communicating physical subsystems. This invention takes advantage of this inherent delay, as described in detail below.
- FIG. 1 presents a pictorial explanation that may aid in understanding the invention. Therein,
vertical lines node 24 is the node of concern, but it is understood that the consideration undertaken bynode 24 are concurrently taken by all other nodes. - The horizontal distances between
line 24 and the other lines represent the time delays for events in other nodes to affect the value or state atnode 24. Accordingly,event 30 processed at node 21 for time T₁ may cause anevent 40 atnode 24 to be scheduled for some time not earlier than T₂, as shown in FIG. 1. The interval between T₁ and T₂ equals the delay betweenline 24 and line 21 (i.e., the horizontal distance between the lines). - The events depicted in FIG. 1 can be divided into two groups: those that have been simulated (and marked with a cross), and those that are yet to be simulated (un-crossed). The simulated events need not be considered for simulation because their effects are already represented by those events which have not been simulated (for example,
event 30 has causedevent 40; the first has been simulated, while the second is yet to be considered for processing). - Of the events that have yet to be simulated (33-43), event 33 in FIG. 1 is earliest in time from among all of the nodes. In this case, the time of event 33, forms the current floor simulation time of the system. The floor is depicted in FIG. 1 by dashed line 45.
- A time interval beginning with the floor simulation time is selected for consideration. This time interval, which I call the bounded lag interval, can be a convenient time interval that takes into account the number of nodes in the system, the number of events to be simulated, and the computing power of the processors employed. All events scheduled within the bounded lag interval can be affected by events scheduled at other nodes within the bounded lag time interval, but only if those nodes are at a time distance from the affected
node 24 that is less than the selected bounded lag interval. That reduces the number of nodes that need to be considered. In the FIG. 1 depiction, the bounded lag interval ends at dashedline 46; and as drawn, the nodes that need to be considered for their potential effect onnode 24 arenodes Nodes 21 and 27 are outside the bounded lag delay and their scheduled events within (or outside) the bounded lag need not be considered. - In considering the effects on
node 24 in greater detail, one can take into account the time of the next scheduled event and reduce the number of nodes being considered even further. In the FIG. 1 embodiment, for instance, the next scheduled event isevent 35, and as drawn, onlynodes - In determining whether
event 35 is to be simulated, one can observe thatonly event 34 is scheduled early enough to have a possible impact onevent 35. That event can be analyzed and if it is determined that it does not affectevent 35, thenevent 35 can be simulated. Alternatively, it may prove even more efficient to refrain from processingevent 35 simply because of the potential effect byevent 34. In the following description, this alternative approach is taken because it saves the process of evaluating whatevent 34 may do. - FIG. 2 presents a block diagram of a concurrent event simulator. It comprises a plurality of
node controllers 51 that are connected to a communications andcommon processing network 52.Network 52 performs the communication and synchronization functions, and in some applications it also performs some of the computations that are needed by the entire system. The node controllers can be realized with conventional stored program computers that may or may not be identical to each other. Each of the node controllers, which corresponds to a node in the FIG. 1 depiction, is charged with the responsibility to simulate a preassigned subsystem of the simulated system. Each controller Ci maintains an event list Πi that is executed by simulating each event in the list in strict adherence to the scheduled event simulation times. It should be remembered that the bounded lag interval is selectively fixed, and that in conformance with the selected bounded lag interval eachcontroller 51 is cognizant of the processors with which it must interact to determine whether events are scheduled. The process by which the event simulations are enabled is carried out by the controllers in accordance with the above-described principles, as shown, for example, by the flowchart of FIG. 3. - The process described in FIG. 3 is carried out in iterations that encompass
blocks 100 through 140.Block 100 is the flow control block. It tests whether the floor simulation time, Tfloor, is less than the end of the simulation time, Tend. As long as Tfloor < Tend, the simulation continues by passing control to block 110. When Tfloor reaches or exceeds Tend, the simulation ends.Block 110 determines the floor simulation time of the simulated system at each iteration. That is, block 110 determines the lowest event time of the scheduled events dispersed among the event lists (Πi) of controllers 51 (Ci) that are are yet to be simulated. Expressed mathematically, block 110 evaluates the floor simulation time in accordance with the equationnode controllers 51 and Ti is the time of the event, ei, which has the earliest scheduled time among the events e' in the event list Πi; i.e.,Block 110 can be implemented in each of the controllers by having each controller broadcast to all other controllers (via network 52) its Ti and evaluate for itself the minimum Ti which defines Tfloor. Alternatively, block 110 can be implemented within communications andcommon processing network 52 by havingcontrollers 51 send their Ti values to network 52, and havingnetwork 52 select the minimum Ti and broadcast it as Tfloor back to the controllers. - Having established the Tfloor, and knowing the system's bounded lag interval, B, that limits the number of neighboring controllers with which a controller must communicate, in accordance with
block 120, each of the controllers evaluates its earliest "at risk" time. This is accomplished bynetwork 52 distributing the Ti information to neighboring controllers, as required, and each controller Ci evaluates the "at risk" demarcation point, αi, from the Ti information. This "at risk" point is defined as the earliest time at which changes at the neighboring controllers can affect the history simulated by the controller, based on the neighboring controllers' own scheduled events or based on a response to an event from the controller itself (reflection). This is expressed by the following equation: - Having determined the value of αi which corresponds to the point in the simulated time beyond which the simulation of events at controller Ci is "at risk", in accordance with
block 130, processor Ci simulates all or some of the scheduled events whose times are earlier that αi. Inblock 140, the time Ti is advanced with each simulation of an event, and the simulated event is deleted from Πi. Concurrently, if in the course of simulating an event, new events are called to be scheduled, then those events are sent to network 52 for transmission to the appropriate node controllers. Similarly, if the execution of events is called to be blocked, that information is also sent to network 52, and thereafter to the appropriate node controllers for modification of the event lists. - The following carries out the example depicted in FIG. 1, assuming for the sake of simplifying the drawing that the situation remains static -- i.e., none of the depicted events are cancelled and no unshown events are scheduled. After
events node 24, the bounded lag defined by the distance between dashed line 45 and dashedline 46 specifies that only nodes 22-26 need to be considered. In the course of that consideration,node 24 determines that scheduledevent 34 atnode 25 defines an "at risk"demarcation point 50. Since there are no events scheduled fornode controller 24 between the time of Tfloor, andpoint 50, no progress in simulations is made by this controller. Concurrently,node controller 22 simulates event 33 (since it is positioned at Tfloor, and no other event can affect it), andnode 25 simulatesevent 34 since neithernode 26 nor node 24 (the closest nodes) have any events scheduled prior to the time ofevent 34.Node 27 probably also simulates event 36, but this is not certain, since FIG. 1 does not show all of the neighbors ofnode controller 27. - With
events 33, 34, and 36 simulated and deleted from their respective event lists, the next iteration raises Tfloor to the time of event 37 (and correspondingly raises the horizon or end of the bounded lag interval. This end is equal to Tfloor+ B). At this new level of Tfloor the "at risk" demarcation point fornode 24 is at point 49 (dictated byevent 39 of node 23), and in accordance with this new "at risk" point, bothevents node controller 24. This completes the simulation of events scheduled fornode 24 which are shown in FIG. 1. Concurrently atnode 23,event 39 is simulated, event 37 atnode 27 is simulated, but atnode 26event 38 is not simulated because it is beyond the "at risk" point ofnode 26 caused by the position of event 37 atnode 27. The next Tfloor moves to the time ofevent 38, and the process continues to simulate additional events. - The above description concentrates on evaluating the "at risk" demarcation point in connection with the direct effects of one node on another. In many physical systems, however, there are many instances where one subsystem affects another subsystem indirectly, i.e., through some other subsystem. This condition gives rise to the possibility that the intermediate node is either busy and unavailable generally, or is somehow sensitized to serve some nodes and not other nodes. Either situation can yield the condition that the delay from one node, A, to another node, C, through an intermediate node, B, is at times much longer than the sum of the of delays A to B and B to C. I call this additional delay an opaque period. Opaque periods have the potential for pushing forward the "at risk" demarcation point and, therefore, it is beneficial to account for this potential in evaluating αi. Such accounting may be achieved by evaluating αi iteratively as follows.
- 1: Set
- 2: synchronize
- 3: evaluate
- 4: evaluate
- 5: synchronize
- 6: evaluate
- 7: if A ≤ Tfloor + B then increment k and return to step 3.
- In the above, the term neighbors(i) refers to nodes that communicate directly with node i. The auxiliary variable β
-
- It may be observed that with each iteration the value of Tfloor increases because the event determining that value can always be simulated. The movement of Tfloor is affected, however, by how closely the nodes are separated and the scheduled events. One other observation that can be made is that the above-described procedure is very conservative, in the sense that an assumption is made that whenever an event is scheduled to be simulated at one node controller, it will always have an effect on the neighboring controller (after the appropriate delay). In physical systems, however, there are many situations where one part of a system performs many operations that affect none of its neighbor subsystems, or affect only one or very few of its neighbor subsystems. Knowledge that an event scheduled for simulation will not affect a neighboring node can be put to use to simulate more events concurrently (have fewer node controllers that are idle). This can be accomplished by communicating not only the Ti of the earliest scheduled event in each list, but also the effect that it may have on neighboring node controllers. If fact, each list can broadcast more than just the earliest scheduled event. The design decision that a practitioner must make, of course, is whether fewer iterations but more complex evaluations are economically justifiable.
- FIG. 4 presents a block diagram of one realization for
node controller 51. Although this embodiment relates to use of the iterative method for evaluating α (with the use of the auxiliary variable β) when the delays between changes in one subsystem and the effect of those changes on other subsystems are not insignificant, it will be appreciated by the skilled artisan that the other realizations for computing αi are substantially similar to this realization. It will also be appreciated that althoughnode controller 51 is shown in FIG. 2 as one of a plurality of controllers, such plurality can be realized within a single processor of sufficient computing power. - In FIG. 4, state register 56 defines the current state of the simulated subsystem and
event list 53 specifies the events that are scheduled to be simulated.Processor 54 is the event simulation processor, and it is responsive to state register 56, toevent list 53 and to register 58.Register 58 is the α register and, as the name implies, it stores the value of α for the controller. Based on the value of α and the scheduled time of the event at the top of the event list,processor 54 simulates the event in conformance with the state of the subsystem and develops a new state of the subsystem as well as, perhaps, new events. The new state is stored inregister 56, new events scheduled for the controller are stored inevent list 53 vialine 61, and events affecting other controllers are communicated out vialine 62. Events produced at other controllers that may affect this controller are accepted vialine 63 and stored inevent list 53 throughprocessor 54. - Whereas
processor 54 is the event simulation processor,processor 55 is the synchronization processor.Processor 54 is shown in FIG. 4 as a separate processor but in practice a single processor may serve the function of bothprocessor 54 andprocessor 55.Processor 55 receives information fromevent list 53 concerning the time of the event inlist 53 that possesses the earliest simulation time. It transmits that information to other controllers via line 64 and receives like information from all other relevant nodes vialine 65. From thatinformation processor 55 develops the value of Tfloor and stores that value inregister 57.Processor 55 also receives βj information from its neighbor controllers (controllers where changes can affect the controller directly) vialine 66, and transmits its own βi values vialine 67. With the aid of Tfloor and the other incoming information,processor 55 performs the iterative computations to develop the values of βprocessor 55 inregisters - It is to be understood that the foregoing descriptions are merely illustrative of my invention and that other implementations and embodiments which incorporate variations from the above are possible.
- For example, the above assigns the task of computing Tfloor to
processor 55. However, it may be preferable to include computing means in the communication and common processing network of FIG. 2 where the Tfloor is computed and distributed to the various node controllers.
Claims (20)
- A method for simulating behavior of a system containing interacting subsystems, where a known minimum delay exists between changes occurring at one of said subsystems and the effects of said changes on others of said subsystems, CHARACTERIZED BY:a plurality of simulation steps, with each such step simulating events occurring at a simulation time in each of said subsystems, restricted to events whose simulation time lies within a preselected time interval beginning with the earliest simulation time of events that are yet to be simulated.
- A method as claimed in claim 1 wherein said step of simulating is followed by a step of advancing said preselected interval in simulated time.
- A method as claimed in claim 1 wherein said step of simulating concurrently performs a plurality of event simulations.
- A method as claimed in claim 1 wherein said step of simulating is preceded by a step of evaluating whether a scheduled event is to be simulated, based on states of or changes occurring at a subset of said subsystems in response to previous simulation steps.
- A method as claimed in claim 4 wherein the subsystems belonging to said subset are dependent on the simulation time span of said preselected interval.
- A method as claimed in claim 4 wherein said subset is controlled by the time span of said preselected interval and said delays between said subsystems.
- A method as claimed in claim 4 wherein said subset includes the subsystems whose delays are within said preselected interval.
- A method as claimed in claim 4 wherein said step of evaluating computes a value of α and said step of simulating simulates events whose simulated time is not greater than α.
- A method as claimed in claim 8 wherein said α for subsystem i, αi, is computed by
- A method as claimed in claim 4 wherein said step of evaluating includes consideration of the condition that the effect of a change in one of said subsystems is blocked from propagating through another one of said subsystems.
- A method as claimed in claim 4 wherein said step of evaluating computes a value of α and said step of simulating simulates events whose simulated time is not greater than α, where α for subsystem i, αi, is the value of α2: synchronize5: synchronize7: if A ≤ Tfloor + B then increment k and return to step 3,where Ti is the time of the event in subsystem i having the earliest simulation time, β
- A method as claimed in claim 4 wherein said step of evaluating computes a value of α and said step of simulating simulates events whose simulated time is not greater than α, where α for subsystem i, αi, is the value of α2: synchronize5: synchronize7: if A ≤ Tfloor + B then increment k and return to step 3,where Ti is the time of the event in subsystem i having the earliest simulation time, β
- A system for performing discrete event simulation of a system having a plurality of interacting subsystems, comprising:a plurality of blocks, each block simulating a preselected subsystem and including (a) an event list associated with said preselected subsystem and (b) means for maintaining information concerning the state of said preselected subsystem, and α, where α is an estimate of the earliest time for which the information in said associated state registers or the events in said associated event list can be modified by neighboring subsystems; and characterized bya controller system for repetitively performing simulation of events in said event lists whose times are earlier than α and re-evaluating the values of Tfloor and α, said controller system comprising a plurality of controllers communicating with said blocks, with each controller repetitively performing simulation of events in said event lists and re-evaluating the value of said α; and means for interconnecting said controllers, wherein said interconnecting means receives a Ti from each of said event lists, where Ti is the simulated time of the event to be simulated in said event list having the earliest simulation time, develops Tfloor which is min (Ti), and returns Tfloor to each of said controllers.
- A system as claimed in claim 14 wherein said controller system comprises a plurality of interconnected controllers communicating with said blocks, with each controller repetitively performing simulation of events in said event lists and re-evaluating the value of said α.
- A system as claimed in claim 14 wherein said controller system comprises an event evaluation processor associated with each of said blocks, connected to its associated event list, its associated means for maintaining state information, and to neighboring blocks, which are blocks that simulate subsystems that directly affect, or can be affected by, said preselected subsystem, where said event processor performs simulation of events in said associated event list and modifies said associated event list and the event lists of said neighboring blocks, as required by said simulation of events;a synchronization processor associated with each of said blocks, connected to its associated event list and means for maintaining α, for developing values of said α; andmeans for interconnecting said plurality of event evaluation processors and synchronization processors.
- A system as claimed in claim 16 wherein said synchronization processor develops a value of α based on values of Tj, from a preselected subset of said subsystems, where Tj is the simulation time of the event in subsystem j having the earliest simulation time.
- A system as claimed in claim 16 where each of said synchronization processors develops a value of α for its block based on a consideration of potential changes in neighbors of its block.
- A system as claimed in claim 18 wherein said consideration is iterative.
- A system as claimed in claim 16 where each of said synchronization processors develops a value of α for its block based on a consideration of opaque periods of its neighbors.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US114369 | 1987-10-28 | ||
US07/114,369 US4901260A (en) | 1987-10-28 | 1987-10-28 | Bounded lag distributed discrete event simulation method and apparatus |
Publications (3)
Publication Number | Publication Date |
---|---|
EP0314370A2 EP0314370A2 (en) | 1989-05-03 |
EP0314370A3 EP0314370A3 (en) | 1991-07-31 |
EP0314370B1 true EP0314370B1 (en) | 1996-01-24 |
Family
ID=22354804
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP88309768A Expired - Lifetime EP0314370B1 (en) | 1987-10-28 | 1988-10-19 | Bounded lag distributed discrete event simulation method and apparatus |
Country Status (6)
Country | Link |
---|---|
US (1) | US4901260A (en) |
EP (1) | EP0314370B1 (en) |
JP (1) | JPH0697451B2 (en) |
CA (1) | CA1294046C (en) |
DE (1) | DE3854935T2 (en) |
IL (1) | IL88146A (en) |
Families Citing this family (61)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE3721719A1 (en) * | 1987-07-01 | 1989-01-12 | Lawrenz Wolfhard | METHOD FOR TESTING A NETWORK BUILDING |
JP2583949B2 (en) * | 1988-03-10 | 1997-02-19 | 松下電器産業株式会社 | Logic simulation method and logic simulation device |
GB8812849D0 (en) * | 1988-05-31 | 1988-07-06 | Int Computers Ltd | Logic simulator |
US5452231A (en) * | 1988-10-05 | 1995-09-19 | Quickturn Design Systems, Inc. | Hierarchically connected reconfigurable logic assembly |
US5109353A (en) * | 1988-12-02 | 1992-04-28 | Quickturn Systems, Incorporated | Apparatus for emulation of electronic hardware system |
US5329470A (en) * | 1988-12-02 | 1994-07-12 | Quickturn Systems, Inc. | Reconfigurable hardware emulation system |
US5369593A (en) | 1989-05-31 | 1994-11-29 | Synopsys Inc. | System for and method of connecting a hardware modeling element to a hardware modeling system |
US5353243A (en) | 1989-05-31 | 1994-10-04 | Synopsys Inc. | Hardware modeling system and method of use |
US5161115A (en) * | 1989-09-12 | 1992-11-03 | Kabushiki Kaisha Toshiba | System test apparatus for verifying operability |
US5081601A (en) * | 1989-09-22 | 1992-01-14 | Lsi Logic Corporation | System for combining independently clocked simulators |
US5375074A (en) * | 1990-01-23 | 1994-12-20 | At&T Corp. | Unboundedly parallel simulations |
US5436846A (en) * | 1990-05-29 | 1995-07-25 | Grumman Aerospace Corporation | Method of facilitating construction of a microwave system by appropriate measurements or determination of parameters of selected individual microwave components to obtain overall system power response |
US7373587B1 (en) * | 1990-06-25 | 2008-05-13 | Barstow David R | Representing sub-events with physical exertion actions |
WO1992000654A1 (en) * | 1990-06-25 | 1992-01-09 | Barstow David R | A method for encoding and broadcasting information about live events using computer simulation and pattern matching techniques |
JPH0464164A (en) * | 1990-07-03 | 1992-02-28 | Internatl Business Mach Corp <Ibm> | Simulation method and device |
US5272651A (en) * | 1990-12-24 | 1993-12-21 | Vlsi Technology, Inc. | Circuit simulation system with wake-up latency |
WO1992014201A1 (en) * | 1991-02-01 | 1992-08-20 | Digital Equipment Corporation | Method for multi-domain and multi-dimensional concurrent simulation using a conventional computer |
JPH05216712A (en) * | 1991-10-23 | 1993-08-27 | Internatl Business Mach Corp <Ibm> | Computer system, method for performing inner-viewing task on computer system and i/o processor assembly |
US5794005A (en) * | 1992-01-21 | 1998-08-11 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Synchronous parallel emulation and discrete event simulation system with self-contained simulation objects and active event objects |
US5633812A (en) * | 1992-09-29 | 1997-05-27 | International Business Machines Corporation | Fault simulation of testing for board circuit failures |
US5913051A (en) * | 1992-10-09 | 1999-06-15 | Texas Instruments Incorporated | Method of simultaneous simulation of a complex system comprised of objects having structure state and parameter information |
US5566097A (en) * | 1993-03-05 | 1996-10-15 | International Business Machines Corporation | System for optimal electronic debugging and verification employing scheduled cutover of alternative logic simulations |
US6363518B1 (en) * | 1993-11-04 | 2002-03-26 | Cadence Design Systems, Inc. | Automated positioning of relative instances along a given dimension |
US5680583A (en) * | 1994-02-16 | 1997-10-21 | Arkos Design, Inc. | Method and apparatus for a trace buffer in an emulation system |
GB2338325B (en) * | 1994-10-03 | 2000-02-09 | Univ Westminster | Data processing method and apparatus for parallel discrete event simulation |
US5617342A (en) * | 1994-11-14 | 1997-04-01 | Elazouni; Ashraf M. | Discrete-event simulation-based method for staffing highway maintenance crews |
US6058252A (en) * | 1995-01-19 | 2000-05-02 | Synopsys, Inc. | System and method for generating effective layout constraints for a circuit design or the like |
US6053948A (en) * | 1995-06-07 | 2000-04-25 | Synopsys, Inc. | Method and apparatus using a memory model |
US5684724A (en) * | 1995-08-30 | 1997-11-04 | Sun Microsystems, Inc. | Flashback simulator |
US5809283A (en) * | 1995-09-29 | 1998-09-15 | Synopsys, Inc. | Simulator for simulating systems including mixed triggers |
US5784593A (en) * | 1995-09-29 | 1998-07-21 | Synopsys, Inc. | Simulator including process levelization |
JP2927232B2 (en) * | 1996-01-29 | 1999-07-28 | 富士ゼロックス株式会社 | Distributed simulation apparatus and distributed simulation method |
US5841967A (en) * | 1996-10-17 | 1998-11-24 | Quickturn Design Systems, Inc. | Method and apparatus for design verification using emulation and simulation |
US6195628B1 (en) | 1997-03-13 | 2001-02-27 | International Business Machines Corporation | Waveform manipulation in time warp simulation |
US5850538A (en) * | 1997-04-23 | 1998-12-15 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Priority queues for computer simulations |
US6389379B1 (en) | 1997-05-02 | 2002-05-14 | Axis Systems, Inc. | Converification system and method |
US6421251B1 (en) | 1997-05-02 | 2002-07-16 | Axis Systems Inc | Array board interconnect system and method |
US6009256A (en) * | 1997-05-02 | 1999-12-28 | Axis Systems, Inc. | Simulation/emulation system and method |
US6134516A (en) * | 1997-05-02 | 2000-10-17 | Axis Systems, Inc. | Simulation server system and method |
US6321366B1 (en) | 1997-05-02 | 2001-11-20 | Axis Systems, Inc. | Timing-insensitive glitch-free logic system and method |
US6026230A (en) * | 1997-05-02 | 2000-02-15 | Axis Systems, Inc. | Memory simulation system and method |
US6031987A (en) | 1997-05-06 | 2000-02-29 | At&T | Optimistic distributed simulation based on transitive dependency tracking |
US5960191A (en) * | 1997-05-30 | 1999-09-28 | Quickturn Design Systems, Inc. | Emulation system with time-multiplexed interconnect |
US6059835A (en) * | 1997-06-13 | 2000-05-09 | International Business Machines Corporation | Performance evaluation of processor operation using trace pre-processing |
US5970240A (en) * | 1997-06-25 | 1999-10-19 | Quickturn Design Systems, Inc. | Method and apparatus for configurable memory emulation |
US6278963B1 (en) | 1997-07-01 | 2001-08-21 | Opnet Technologies, Inc. | System architecture for distribution of discrete-event simulations |
CA2295983A1 (en) * | 1997-07-01 | 1999-01-14 | Mil 3, Inc. | System architecture for distribution of discrete-event simulations |
JP3162006B2 (en) | 1997-11-10 | 2001-04-25 | 核燃料サイクル開発機構 | Simulation method of extraction system |
JP3746371B2 (en) * | 1998-04-09 | 2006-02-15 | 株式会社日立製作所 | Performance simulation method |
US6356862B2 (en) * | 1998-09-24 | 2002-03-12 | Brian Bailey | Hardware and software co-verification employing deferred synchronization |
DE19921128A1 (en) * | 1999-05-07 | 2000-11-23 | Vrije Universiteit Brussel Bru | Process for generating target system-specific program codes, simulation processes and hardware configuration |
US7257526B1 (en) * | 2000-04-05 | 2007-08-14 | Lucent Technologies Inc. | Discrete event parallel simulation |
US20020133325A1 (en) * | 2001-02-09 | 2002-09-19 | Hoare Raymond R. | Discrete event simulator |
US7219047B2 (en) * | 2001-03-29 | 2007-05-15 | Opnet Technologies, Inc. | Simulation with convergence-detection skip-ahead |
US7315805B2 (en) * | 2004-02-05 | 2008-01-01 | Raytheon Company | Operations and support discrete event stimulation system and method |
US7502725B2 (en) * | 2004-04-29 | 2009-03-10 | International Business Machines Corporation | Method, system and computer program product for register management in a simulation environment |
US20070130219A1 (en) * | 2005-11-08 | 2007-06-07 | Microsoft Corporation | Traversing runtime spanning trees |
US20080270213A1 (en) * | 2007-04-24 | 2008-10-30 | Athena Christodoulou | Process risk estimation indication |
JP5251586B2 (en) * | 2009-02-19 | 2013-07-31 | 富士通セミコンダクター株式会社 | Verification support program, verification support apparatus, and verification support method |
WO2018054465A1 (en) * | 2016-09-22 | 2018-03-29 | Siemens Aktiengesellschaft | Method and devices for the synchronised simulation and emulation of automated production systems |
DE102018111851A1 (en) * | 2018-05-17 | 2019-11-21 | Dspace Digital Signal Processing And Control Engineering Gmbh | Method for event-based simulation of a system |
Family Cites Families (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4204633A (en) * | 1978-11-20 | 1980-05-27 | International Business Machines Corporation | Logic chip test system with path oriented decision making test pattern generator |
US4527249A (en) * | 1982-10-22 | 1985-07-02 | Control Data Corporation | Simulator system for logic design validation |
JPS59119466A (en) * | 1982-12-27 | 1984-07-10 | Fujitsu Ltd | Inter-unit communication system of hardware logic simulator |
JPS59163645A (en) * | 1983-03-09 | 1984-09-14 | Hitachi Ltd | Function simulating method |
GB8309692D0 (en) * | 1983-04-09 | 1983-05-11 | Int Computers Ltd | Verifying design of digital electronic systems |
US4636967A (en) * | 1983-10-24 | 1987-01-13 | Honeywell Inc. | Monitor circuit |
US4751637A (en) * | 1984-03-28 | 1988-06-14 | Daisy Systems Corporation | Digital computer for implementing event driven simulation algorithm |
DE3445030A1 (en) * | 1984-12-11 | 1986-06-19 | Philips Patentverwaltung Gmbh, 2000 Hamburg | TRAFFIC SIMULATION DEVICE FOR TESTING SWITCHING SYSTEMS WITH CONSIDERATION OF THE SUBSCRIBER-SYSTEM INTERACTION |
JPS61237162A (en) * | 1985-04-15 | 1986-10-22 | Mitsubishi Electric Corp | Event simulating system |
-
1987
- 1987-10-28 US US07/114,369 patent/US4901260A/en not_active Expired - Fee Related
-
1988
- 1988-10-19 DE DE3854935T patent/DE3854935T2/en not_active Expired - Fee Related
- 1988-10-19 EP EP88309768A patent/EP0314370B1/en not_active Expired - Lifetime
- 1988-10-24 CA CA000581027A patent/CA1294046C/en not_active Expired - Lifetime
- 1988-10-25 IL IL88146A patent/IL88146A/en not_active IP Right Cessation
- 1988-10-28 JP JP27105688A patent/JPH0697451B2/en not_active Expired - Fee Related
Also Published As
Publication number | Publication date |
---|---|
JPH0697451B2 (en) | 1994-11-30 |
JPH01155462A (en) | 1989-06-19 |
IL88146A0 (en) | 1989-06-30 |
DE3854935D1 (en) | 1996-03-07 |
EP0314370A2 (en) | 1989-05-03 |
US4901260A (en) | 1990-02-13 |
EP0314370A3 (en) | 1991-07-31 |
IL88146A (en) | 1992-03-29 |
CA1294046C (en) | 1992-01-07 |
DE3854935T2 (en) | 1996-09-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP0314370B1 (en) | Bounded lag distributed discrete event simulation method and apparatus | |
Righter et al. | Distributed simulation of discrete event systems | |
Peacock et al. | Distributed simulation using a network of processors | |
Heidelberger et al. | Computer performance evaluation methodology | |
Bailey et al. | Parallel logic simulation of VLSI systems | |
Bagrodia et al. | MIDAS: Integrated design and simulation of distributed systems | |
Madisetti et al. | WOLF: A rollback algorithm for optimistic distributed simulation systems | |
Cota et al. | A modification of the process interaction world view | |
Bagrodia et al. | A unifying framework for distributed simulation | |
Kurose et al. | Computer-aided modeling, analysis, and design of communication networks | |
Fromm et al. | Experiences with performance measurement and modeling of a processor array | |
Hsu et al. | Performance measurement and trace driven simulation of parallel CAD and numeric applications on a hypercube multicomputer | |
Ayani | Parallel simulation | |
Abrams et al. | Implementing a global termination condition and collecting output measures in parallel simulation | |
Reed | Parallel discrete event simulation: a case study | |
Tay et al. | Performance analysis of time warp simulation with cascading rollbacks | |
Kumar et al. | A study of achievable speedup in distributed simulation via null messages | |
MacNair | An introduction to the research queueing package | |
Smith et al. | Performance analysis of software for an MIMD computer | |
Chiola et al. | Exploiting timed petri net properties for distributed simulation partitioning | |
Peterson et al. | Performance of a globally-clocked parallel simulator | |
Theodoropoulos et al. | Analyzing the Timing Error in Distributed Simulations of Asynchronous Computer Architectures. | |
Binns et al. | An implementation for hybrid continuous variable/discrete event dynamic systems | |
Gimarco | Distributed simulation using hierarchical rollback | |
Zarei et al. | Performance analysis of automatic lookahead generation by control flow graph: some experiments |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
AK | Designated contracting states |
Kind code of ref document: A2 Designated state(s): DE GB |
|
PUAL | Search report despatched |
Free format text: ORIGINAL CODE: 0009013 |
|
AK | Designated contracting states |
Kind code of ref document: A3 Designated state(s): DE GB |
|
17P | Request for examination filed |
Effective date: 19920122 |
|
17Q | First examination report despatched |
Effective date: 19940228 |
|
RAP3 | Party data changed (applicant data changed or rights of an application transferred) |
Owner name: AT&T CORP. |
|
GRAA | (expected) grant |
Free format text: ORIGINAL CODE: 0009210 |
|
AK | Designated contracting states |
Kind code of ref document: B1 Designated state(s): DE GB |
|
REF | Corresponds to: |
Ref document number: 3854935 Country of ref document: DE Date of ref document: 19960307 |
|
PLBE | No opposition filed within time limit |
Free format text: ORIGINAL CODE: 0009261 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT |
|
26N | No opposition filed | ||
REG | Reference to a national code |
Ref country code: GB Ref legal event code: 732E |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: GB Payment date: 20000914 Year of fee payment: 13 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: DE Payment date: 20001221 Year of fee payment: 13 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: GB Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20011019 |
|
REG | Reference to a national code |
Ref country code: GB Ref legal event code: IF02 |
|
GBPC | Gb: european patent ceased through non-payment of renewal fee |
Effective date: 20011019 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: DE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20020702 |